Today, however, the picture is totally different. The current models are an almost endless gold mine of insight into both how to build AI well and what can sometimes go wrong with it if it isn’t built well. I believe that if slowing down bought us even an extra year or two before models reach critical levels of capability, and we used that time to advance alignment, we could greatly reduce the risk that something goes seriously wrong.
When applying the Utilitarian lens, Amodei points out how it would be beneficial to slow the production of AI so that we can better align with it so that there are safety protocols in place. This way would cause the least harmful affect because if AI companies were to slow down to better understand all aspects of what they are creating, they can better ensure that they will not get any unwanted or unfriendly outcome from their AI creations. This helps to keep everyone safe while still advancing the AI field, albeit a little slower than it is currently being developed. With that, while it may be beneficial to society in order to ensure its safety, it does come with the pitfall that it will take longer to find solution to different problems that they are using AI for which could cause harm along the way.